1. Field of the Invention
The invention relates generally to computer software metrics and, more specifically, to defining and utilizing a metric that provides an objective, quantitative method of evaluating the physical effort required to manipulate a computer-human-interface.
2. Description of the Prior Art
The prior art employs various techniques to design a computer-human interface. Among these are subjective techniques which utilize experience and intuition to appraise existing or proposed interface designs and generally result in lists of specific recommendations that have historically been found integral to good interface design (Mayhew, Deborah, Principles and Guidelines in Software User Interface Design, 1992), (Smith, S. L. and Mosier, J. M., Design Guidelines for User Interface Software, 1986). However, it is not always apparent that resulting recommendations are applicable to new technologies since, being based on phenomenologic observation, the scope of their applicability is uncertain.
Less subjective approaches to interface design comprise such procedures as prototyping, focus groups, cognitive walk-throughs, and alpha tests. A review of writings of those skilled in these arts (Dix, A., Finley, J., Aboud, G., Beale, R., Human-Computer Interaction, 1993), (Shneiderman, B., Designing the User Interface: Strategies for Effective Human-Computer Interaction, 1992); (Dumas, J., Designing User Interfaces for Software, 1988) convey that similarities exist between different practitioners. In general, these experts advocate: (1) employ practical experience, (2) use applicable experimental findings, (3) use rules-of-thumb, (4) promote consistency. Attempts to apply practical experience and rules-of-thumb confront the analyst with the same quandaries of observed with the subjective approaches. Although experimental results can be of significant worth, implications of the artificial setting under which experiments are typically conducted must be appraised for applicability to the production environment anticipated. Finally, the provision of consistency, while widely advocated, has proved difficult to achieve since an objective, quantitative definition of consistency has yet to be formulated. (Payne S. J. and Green, T. R. G., Task-Action Grammars: A model of the mental representation of task languages, Computer-Human Interaction, 1986, pp. 93-133)
Dumas and Redish (Dumas, J. S. and Redish, J. C. A Practical Guide to Usability Testing, 1993), (Hix, Deborah and Hartson, H. Rex, Developing User Interfaces: Insuring Usability Through Product and Process, 1993) summarizes the prior art for comparing existing interfaces: (1) provide expert review of the design, (2) perform peer walk-through, (3) prototype, and (4) monitor behavior of the production environment via one-way mirrors, logs, video tapes, and questionnaire interviews. Of these, expert review has been shown inadequate when performed by a single expert but of benefit when performed by a group of outside experts (Jeffries, R., Miller, J. R., Wharton, C., and Uyed, C. M., User Interface Evaluation in the Real World, Proceedings of CHI '91, Human Factors in Computing Systems, 1991, pp. 119-124.). While frequently employed to advantage, peer walk-through and prototyping have the disadvantage of producing difficult to reproduce, subjective evaluations. The monitoring of production environments captures extensive data, but such data only offer the opportunity of objective quantification; there remains the need to define the methodology for converting such data into recognized, quantitative measures that accurately reflect usability.
Since it predicts the time required to acquire a target, some experts proffer Fitts' Law as the basis for an objective, quantitative metric of the effort expended during physical actions that acquire a stationary target. An expression of Fitts' Law is: ##EQU1## where: TT Total time for target acquisition (seconds).
RT Reaction time (seconds). PA2 b Muscle transfer rate (seconds per bit). PA2 I Index of Difficulty (bits). PA2 W.sub.t Width of target (linear units). PA2 D.sub.t Distance to target (linear units).
This invention relates to the implications of how users perceive the D.sub.t and W.sub.t parameters of the Index of Difficulty: ##EQU2## Fitts (Fitts, P. M., The information capacity of the human motor system in controlling amplitude, Journal of Applied Psychology, 1954, pp. 381-391) implicitly set k=0 by excluding k from his original formulation. Welford (Welford, A., The Fundamentals of Skill, 1968) proposed k=0.5, arguing that this offers a superior fit to published empirical data. MacKenzie (MacKenzie, I., Fitts' Law as a research and design tool in human-computer interaction, Human-Computer Interaction, 1992, pp. 91-139) reappraised the use of Shannon's Information theory (Shannon and Weaver, The Mathematical Theory of Communication, 1949) as the basis of Fitts' Law and concluded that k=1 offers an even better fit to published empirical data. Analysis presented below in the section "Analysis of Arbitrary Triangular Targets" shows that k is not relevant to the user's determination of either target distance or target width.
Studies (Welford, op. cit.), (Jagacinski, R. J. and Monk, D. L., Fitts' Law in two dimensions with hand and head movements, Journal of Motor Behavior, 1985, pp. 77-95); (Jagcinski, R. J., Repperger, D. W., Moran, M. S., Ward, S. L., and Glass, B., Fitts' Law and the microstructure of rapid discrete movements, Journal of Experimental Psychology, 1980, pp. 309-320.) show that times for hand-homing in either direction between keyboard and mouse conform to Fitts' Law. It has also been shown that finger targeting between keys follows Fitts' Law when 7W.sub.t .ltoreq.D.sub.t. Experimental evidence (Card, S. K., English, W. K., and Burr, B. J., Evaluation of the mouse, rate-controlled isometric joy stick, step keys, and text keys for text selection on a CRT, Ergonomics, 1978, pp. 601-613); (Jagacinski and Monk, op. cit.); (MacKenzie, op. cit.) also indicates that Fitts' Law can be applied to CRT environments that employ the mouse controlled cursor. Verification of the applicability of Fitts' Law to mouse movement in real world systems in general requires additional study since cursor control experiments typically employ circular or square targets rather than the more diverse target shapes found in production environments.
The present invention contends that failure of practitioners of the prior art to successfully offer a defensible definition for the D.sub.t, and W.sub.t parameters of Fitts' Index of Difficulty that is applicable to arbitrary targets has rendered its general application unfeasible. Investigating implications of this failure starts by recalling that the aspect ratio (AR) of a target is defined as ##EQU3## Since persons utilizing Fitts' Law generally define W.sub.t of circular and square targets to be the diameter and the length of a side respectively, studies which employ the prior art generally apply to physical targets for which unitary aspect ratios pertain. For such targets the prior art generally defines, D.sub.t, to be the distance from the cursor to the target center although some practitioners of the prior art question the applicability of these definitions of D.sub.t and W.sub.t for targets of 0&lt;AR&lt;&lt;1 and 1&lt;&lt;AR&lt;.infin.. (Gillan, D., Holden, K., Adam, S., Rudisill, M., Magee, L., How Does Fitts' Law Fit Pointing and Dragging, Proceedings of CHI '90; Human Factors in Computing Systems, 1990, pp. 175-182). FIG. 1 conveys implications of these definitions. Parts A through C of FIG. 1 depict rectangular targets of AR=1, AR&gt;1, and AR&lt;1 respectively, while Part D depicts a convex polygon of triangular shape. For these targets the prior art generally defines locations 1A08, 1B08, and 1C08 as the terminus of user traverses into respective physical targets. Under definitions prevailing in the prior art target distances, D.sub.t, are depicted by ##EQU4## for the said targets respectively. For the square target, 1A02, the width and distance definitions are specific, but characteristics of the remaining targets do not permit this specificity for either width or distance definitions. Location 1D08 is not so appraised as triangular targets are not generally covered by studies investigating Fitts' Law.
Defining a constrained target to be a target for which the user considers all target sides when choosing a traverse to the target, research involved with the present invention concludes on logical grounds that traditional definitions of target width and distance are valid only for constrained squares and circles. Data collected and analyzed during development of the present invention provide experimental results showing there are statistically significant differences between mean target acquisition points and centers of targets of the types depicted by FIG. 1 Parts B through D. These empirical results indicate that users confronting interface environments depicted by FIG. 1 Parts B through D with initial cursor locations at 1B18, 1C18, 1D18 will take traverses 1B16, 1C16, and 1D16 and generally have modal termination points at 1B06, 1C06, and 1D06 respectively rather than traverses 1B14, 1C14, and 1D14 with termination points at 1A08, 1B08, and 1C08 as suggested by the prior art. It is shown below that spatial equivalence between 1A06 and 1A08 arises from the geometry of a small, square target and not from predictive ability of the prior art.
In seeking definitions which better reflect user behavior during target acquisition, MacKenzie (op. cit.) offers the following: (1) the perimeter definition: W.sub.t =H+L, (2) the area definition: W.sub.t =H.times.L, and (3) the angle-of-approach definition: ##EQU5## The perimeter definition implies that a square target of H=L=3 and an elongated rectangular target of H=0.5 and L=5.5 have equal W.sub.t. Under the area definition, a square of H=L=3 and an elongated rectangular target of H=0.5 and L=18.0 also have equal W.sub.t. MacKenzie does not offer theoretical justification or empirical evidence of why such diverse shapes have equivalent Wt values nor does he indicate how to determine D.sub.t.
The angle-of-approach definition specifies target width to be the length of the segment subtended by parallel sides of the target when the traverse is extended through the target, the midpoint of this segment being the traverse terminus. Employing Information Theory, research involved with the present invention shows that rational users facing a rectangular target constrained in the narrow dimension will seek an acquisition point on the axis-of-symmetry; namely, line 216 of FIG. 2. The research conducted for this invention empirically corroborates this conclusion for constrained targets. Lines 204 and 220 of FIG. 2, represent limits to user targeting since traverses terminating to the left of 204 or to the right of 220 are to locations which entail decreasing W.sub.t and either a decreasing D.sub.t or a D.sub.t increasing less rapidly than Wt decreases. For either case traverses into these areas result in an increase of the index of difficulty. It follows that under the angle-of-approach definition, the rational user will terminate a traverse to a rectangular target on line 216 between locations 218 and 240.
Given these conclusions, it is possible to compare the effort of traversing two arbitrary paths from initial cursor location 212 to the line segment defined by points 218 and 240 of FIG. 2. Thus, compare path CUMV having ##EQU6## with path CRNS having ##EQU7## The effort of acquiring the target using point 236 as the acquisition point is: ##EQU8## It is thus logically concluded that under the angle-of-approach hypothesis traverses terminating on the line connecting points 230 and 240 will be perceived by users as requiring equal physical effort.
To determine whether users show indifference to the traverse taken to acquire a rectangular target of non-unitary aspect ratio, the research performed for the present invention included an experiment in which subjects completed repeated trials to such a target within a fixed environment. Analysis of data from this experiment investigated the footprint of hits generated to ascertain whether hits were randomly distributed along the line connecting points 218 and 240 as would be expected if all traverses entail equal physical effort. Hits were found to be non-randomly distributed along said line segment, instead having a modal location related to the initial position of the cursor. Therefore, research involved with the present invention concludes the angle-of-approach for target width can be rejected on both logical and empirical grounds.
Those experienced in the art of design and evaluation of computer-human interfaces thus recognize that the prior art is inadequate to: (1) express objectively each operation users perform on computer input devices during a terminal session, (2) identify the Distance and Width parameters of Fitt's Index of Difficulty for other than square or circular targets, and (3) specify in an objective, quantitative manner the physical effort of performing a specified set of computer tasks.
Accordingly it is an objective of the present invention to provide an InterFace Grammar capable of recording all physical operations users perform on a computer-human interface during a terminal session in a manner to permit subsequent quantitative analysis. It is another objective to provide a defensible method for determination of the distance and size of the implicit target contained within triangular and convex quadrilateral targets of arbitrary size, location, and orientation relative to a target acquiring entity. Finally, it is another objective to provide a method for aggregating the effort expended in acquiring individual targets into an index suitable for identification of that interface which requires lowest total physical effort for performance of a given task set.